Anterior and posterior waves of cyclic her1 gene expression are differentially regulated in the presomitic mesoderm of zebrafish

Somites are transient segmental structures that are formed along the anterior-posterior axis of the vertebrate embryo (for a review, see Saga and Takeda, 2001; Maroto and Pourquié,2001). They are generated from the mesenchymal presomitic mesoderm(PSM), which flanks the notochord on both sides. There are three major phases of somitogenesis. First, the prepatterning of the unsegmented PSM and the establishment of the rostrocaudal polarity of the future somite(Stern and Keynes, 1987; Aoyama and Amasoto, 1988);second, the formation of the somitic border and third, the differentiation of the somites to generate the muscles and vertebrae of the trunk and tail(Tam and Trainor, 1994).

It has long been speculated that the prepatterning of the somites is achieved by an oscillator mechanism in the PSM(Cooke and Zeeman, 1976; Meinhardt, 1986) (reviewed by Dale and Pourquié,2000). The first evidence for this oscillator mechanism was provided by the identification of the c-hairy1 gene (Palmerim et al.,1997), which is dynamically expressed in the PSM of chicken. Owing to its cyclic expression, which progresses from the posterior to the anterior PSM,the cells in the chick embryo undergo several on and off phases of c-hairy1 transcription before they become a somite. c-hairy1encodes a bHLH transcription factor, which is a homologue of the Drosophila pair-rule gene hairy(Ish-Horowicz et al., 1985). More recently, several hairy (h) and Enhancer of split(E(spl)) related genes have been identified, which also have a dynamic expression in the vertebrate PSM. This includes the c-hairy2 and c-Hey2 genes in chick (Jouve et al., 2000; Leimeister et al.,2000) as well as the Hes1 and the Hes7 genes in mouse (Jouve et al., 2000; Bessho et al., 2001a; Bessho et al., 2001b). In zebrafish, nine h/E(spl) related genes have been discovered so far[her1-her6 (von Weizsäcker, 1994; Müller et al., 1996; Pasini et al., 2001), her7: AF240772; her8a/b: AY007990/AY007991 and her9(Leve et al., 2001)] but only two of them, her1 and her7, show an oscillating expression in the PSM (Holley et al.,2000; Oates and Ho,2002). The analysis of a deletion mutant for her1 and her7 as well as morpholino-oligonucleotide (MO) knockdown studies suggest that Her1 and Her7 protein function is required for the prepatterning of the zebrafish PSM (Henry et al.,2002; Oates and Ho,2002). The loss of Her1/Her7 protein leads to somites that show alternating weak and strong boundaries(Henry et al., 2002). In addition, a disruption of rostrocaudal polarity within the somites has been observed (Henry et al., 2002; Oates and Ho, 2002).

The current data suggest that the Delta-Notch signalling pathway is the major trigger of cyclic gene expression in the vertebrate PSM (for reviews, see Maroto and Pourquié,2001; Saga and Takeda,2001). In zebrafish, mutants of deltaD (after eight) and Notch1 (deadly seven), as well as MO-knockdown of deltaC abolish the cyclic expression of her1and her7 (Dornseifer et al.,1997; van Eeden et al.,1998; Takke and Campos-Ortega,1999; Holley et al.,2000; Holley et al.,2002; Oates and Ho,2002). In these cases her1 and her7 usually only show an irregular expression in the anterior PSM and a weak, diffuse expression in the posterior part of the PSM and the tailbud. Furthermore, her1 and her7 appear to crossregulate each other, and both are required for the transcription of deltaC and deltaD(Holley et al., 2002; Oates and Ho, 2002; Henry et al., 2002).

We have focussed here on a better understanding of the differential roles of her1 and her7 in regulating the cyclic gene expression by analysing the effects of MO-knockdown on each other's expression. We have found different roles for Her1 and Her7 in regulating the anterior and posterior parts of cyclic her gene expression. Analysis of the her1 promoter reveals that these anteroposterior differences are the result of separable regulatory elements.

By screening a zebrafish genomic library (MoBiTec) with a her1cDNA probe, one clone containing the upstream region and the first exon as well as several clones containing different downstream parts were isolated. The presence of the her7 gene on the upstream region clone was deduced from the genomic sequence and confirmed by an EST sequence (EST-ID:fb97b02) found in the zebrafish EST database(Sprague et al., 2001). The cDNA clone was obtained from RZPD (clone ID: MPMGp609E0421Q8). Sequences were determined on an ABI377XL sequencer (Perkin Elmer/Applied Biosystems) and submitted to GenBank under the accession numbers AF292032 (genomic sequence for her1 and her7) and AF240772 (her7 cDNA sequence).

Amino acid sequences were aligned using the Pileup program of the GCG software package (Devereux et al.,1984; Senger et al.,1998). Similarity trees were generated using PAUP, calculations are based on Pileup alignments. Trees were displayed using Treeview(Page, 1996). The accession numbers of the compared genes are: c-hairy1: AF032966,c-hairy2 (Jouve et al.,2000), Drosophila hairy: X15905, her1:X97329, her4: X97332, her6: X97333, her7: AF240772, her9: AF301264, mouse Hes1: NM008235, mouse Hes7:AB049065, X-hairy1: U36194, X-hairy2A: AF383159, human HES1: Q14469, human HES7: NM032580.

Fish were bred at 28.5°C in a 14 hour light/10 hour dark cycle. Embryos were collected by natural spawning and staged according to Kimmel et al.(Kimmel et al., 1995). For automated in situ hybridisations we followed the protocol of Leve et al.(Leve et al., 2001) using a programmable liquid handling system described by Plickert et al. (Plickert et al., 1997). The hybridisation temperature for the her1 intron probe had to be reduced to 50°C, because of its high AT content. Digoxigenin- or fluorescein-labelled RNA probes were prepared using RNA labelling kits(Roche). Staining was performed with BM purple (Roche) for single in situ hybridisations or, for double fluorescence in situ hybridisations, Vector®Red Alkaline Phosphatase Substrate Kit I (Vector Laboratories) and the ELF®-97 mRNA In Situ Hybridisation Kit (Molecular Probes) were used according to the method of Jowett and Yan(Jowett and Yan, 1996). Whole-mount embryos were observed under a stereomicroscope (Leica) and digitally photographed (Axiocam, Zeiss). Flat-mounted embryos were observed with an Axioplan2 microscope (Zeiss). For observation of the Vector® Red staining or the ELF®-97 precipitate a rhodamine filter set or a DAPI filter set was used, respectively.

A 10.9 kb NcoI fragment from the upstream region of the her-1 gene was in-frame subcloned into the start methionine of the coding sequence of pEGFP (Clontech). After digestion with PstI and EcoRI, the promoter-reporter construct containing 8.6 kb upstream sequence and the EGFP reporter was inserted into pHSREM1 (acc. no. ATCC37642,kindly provided by D. Knipple) to yield construct I. Transgenic lines were produced by injection of a PstI-linearized fragment of this construct, or of PCR-amplified promoter deletions of it, into single cell embryos. For PCR-amplification the Expand High Fidelity PCR System (Roche) was used. For generation of constructs II-VI the same downstream primer (M13for:5′-GTA AAA CGA CGG CCA GT-3′) was used in combination with upstream primer II (5′-TAA ACT TTC CCC AGT CAG-3′), upstream primer III (5′-AAA GCC ACA TCA AAG CCC-3′), upstream primer IV(5′-TTA GCC ATG AAC GAT GCC-3′), upstream primer V (5′-AGC AAC TCC ATA AAA TCC-3′), upstream primer VI (5′-CTA TGA GAC AAC GAT GAG-3′), respectively. Between five and 15 transgenic lines were obtained in each case, but not all showed a sufficiently strong expression. Four, two and one line were eventually analysed for constructs I-III. There were only quantitative (expression level) but no qualitative differences between the lines. The DNA fragments were gel purified prior to injection using a gel extraction kit (Bio-Rad, Gibco BRL). The DNA concentration of the injected solutions was between 80-100 ng/μl in water containing 0.2% phenol red and 0.1 M KCl. Injections were carried out using FemtoJet® and a Micromanipulator (Eppendorf). To test for possible transgenic animals, DNA of 100 embryos was extracted as described previously(Meng et al., 1999). Positive PCR controls were the Wnt5a sense primer 5′-CAG TTC TCA CGT CTG CTA CTT GCA-3′ and the Wnt5a antisense primer 5′-ACT TCC GGC GTG TTG GAG AAT TC-3′. For the transgene test, two GFP primers were used (GFPfor3:5′-CGG CAA CTA CAA GAC CCG CG-3′ and GFPrev3: 5′-GTC CTC GAT GTT GTG GCG GA-3′). The following PCR profile was carried out: 95°C for 2 minutes, then 35 cycles 95°C for 15 seconds, 55°C for 15 seconds and 72°C for 30 seconds followed by an elongation step of 72°C for 5 minutes after the 35 cycles.

Antisense morpholino-modified oligonucleotides (GeneTools) were designed against the first 25 nucleotides of the 5′-UTR and against the start of the ORF of both the her1 cDNA (X97329) and the her7 cDNA(AF240772). Sequence for her1-anti5′ morpholino: 5′-AGT ATT GTA TTC CCG CTG ATC TGT C-3, sequence for the her1-antiATG morpholino: 5′-CAT GGC TGA AAA TCG GAA GAA GAC G-3′, sequence for her7-anti5′ morpholino: 5′-ATG CAG GTG GAG GTC TTT CAT CGA G-3′, sequence for the her7-antiATG morpholino: 5′-CAT TGC ACG TGT ACT CCA ATA GTT G-3′. 0.5 mM of the her7^mos and 1 mM of the her1^mos were injected into single-cell stage embryos. The injection solution additionally contained 0.1 M KCl and 0.2%phenol red. Control injections were done with the morpholino-modified oligonucleotide recommended by GeneTools, or with buffer. The death rate caused by injection of the different morpholinos was usually between 5 and 11%.

The genomic region of the her1 gene was analysed by sequencing two overlapping genomic clones. The her1 gene consists of 4 exons distributed over approximately 6 kb (Fig. 1). We found that her7 is located in a head to head orientation approximately 11 kb upstream of her1. A her7cDNA (acc. no. AF240772) and her1 were previously mapped at 121.5 cM on linkage group 5 (Kelly et al.,2000). All of the other known her genes are located on different chromosomes [see ZFIN: www.zfin.org(Sprague et al., 2001)].

Alignment and phylogenetic analysis of her1 and her7shows that the gene products are only distantly related to each other and are also not closely related to the oscillating c-Hairy1/2 genes from chicken or mouse Hes1 (Fig. 2 and see supplemental data). The Her7 protein, like the Her1 protein, shows all structural features of a subfamily of basic Helix-Loop-Helix (bHLH) proteins that act as transcriptional repressors. This includes a conserved proline in the basic domain, two additional helices termed the orange domain(Dawson et al., 1995) and a WRPW motif (Fisher et al.,1996) at the C terminus. One feature by which the Her7 protein can be distinguished from the other h/E(spl) proteins is a proline residue occurring C-terminally of the WRPW motif. The Her7 protein has this residue in common with the human and mouse Hes7 as well as with the enhancer of split related proteins ESR-4/5 from Xenopus(Jen et al., 1997; Bessho et al., 2001a).

The spatiotemporal expression pattern of her7 mimics that of her1 (Müller et al.,1996; Oates and Ho,2002) (see also supplemental data). To analyse whether her1 and her7 are expressed in the same cycle phase, we have used double fluorescent in situ hybridisation, a technique that avoids detection interference(Jowett and Yan, 1996). Superimposition of the signals shows that her7 expression overlaps exactly that of her1 in the tailbud and the posterior two stripes in the presomitic mesoderm (Fig. 3). In the anterior-most stripe, her7 expression appears weaker than her1 expression, which may be due to faster degradation. In a previous study, her1 expression has been found to be more persistent at the posterior borders of each stripe than her7(Oates and Ho, 2002). This discrepancy could be due to the different techniques that were employed.

Given the overlapping expression of her1 and her7, we have investigated whether they might play similar roles in regulating cyclic gene expression. Antisense morpholino-modified oligonucleotides (MO) against her1 and her7 were used to specifically inhibit the function of these genes. All effects described below were observed with a high penetrance (>90%) in at least 75-175 embryos and in several independent experiments using two different MOs in each case, and were not observed in control injections (see Fig. 4legend for details). We find that in both, the her1^mo- and the her7^mo-injected embryos (or morphants) their own stripe expression is equally strongly perturbed(Fig. 4H-L and O-S) and knockdown of both bHLH genes lead to a similar disruption of deltaCand deltaD expression patterns(Fig. 4M,N,T,U). Comparable effects have been shown previously with different MO variants(Holley et al., 2002; Oates and Ho, 2002). However,there are some important differences in details that we want to discuss in turn.

The analysis of MO-mediated disruption of a gene function on its own expression needs to take into account that the MO can stabilise the mRNA, even though the mRNA is not translated (Oates and Ho, 2002). We have therefore analysed the her1expression pattern both with an exon probe and an intron probe to differentiate between transcriptional regulation and mRNA stability effects. For riboprobe production we have used a part of the largest intron of her1, which yielded in wild-type embryos the same expression pattern as a her1 exon probe, albeit much weaker(Fig. 4A-C, note the different staining times). The exon probe shows a broad and strong expression in the PSM of her1 morphants (Fig. 4H) (Holley et al.,2002; Oates and Ho,2002). The intron probe, however, reveals a different expression pattern and therefore confirms a suspected MO-mediated stabilisation effect. We found embryos with a distinct posterior U-shaped domain with a broad stripe at the anterior end (Fig. 4I)and embryos with an anterior stripe separated from the posterior domain(Fig. 4J). These two types of patterns occur in nearly equal frequencies (71/75 out of 146 embryos analysed), comparable to the pattern of the cyclic expression of the U-shaped domain in wild-type embryos. Thus it appears that only the two most anterior PSM stripes are not formed, while the posterior expression is less affected and apparently oscillates. This changes the previous inference that her1 acts as a repressor of itself(Holley et al., 2002; Oates and Ho, 2002), and suggests instead that it is required as an activator for its own two anterior-most PSM stripes. Monitoring the expression of her7 in her1 morphants shows that her7 is strongly affected in the same way as her1 (Fig. 4K,L), while in a previous study only mild effects on cyclical behaviour could be observed (Oates and Ho,2002), which are probably caused by insufficient penetrance of the morpholino used. Thus, for Her1 we find a common role in activating the her1 and her7 stripe formation in intermediate and anterior PSM. In addition, Her1 is not involved in the her1 and her7wave generation in the posterior PSM, neither as a repressor nor as an activator.

Embryos injected with the her7^mo also show disruption of her1 and her7 cycling, but with different effects on each and differences when compared to her1 morphants. her1 is expressed in a punctate pattern in the PSM of her7 morphants and a more pronounced domain is observed in the tailbud(Fig. 4O,P,Q). her7also shows a pronounced tailbud expression in her7 morphants, and at best a residual weak signal without punctate pattern in the intermediate PSM(Fig. 4R,S). Thus, it appears that the mRNA stabilisation effect is much weaker than that for her1. However, the perturbation of her1 and her7 wave generation in posterior PSM of the her7 morphants, which is not seen in her1 morphants, indicates a unique role for the Her7 protein in this process.

The above results indicate that at least her1 regulation is governed by two distinct phases. This is supported by reporter gene constructs in transgenic lines. her1 upstream fragments of various size were fused to GFP as reporter and the DNA was injected into early embryos. Stably transformed lines were established from these and analysed for their regulatory effects by in situ hybridisation to the GFP mRNA. Lines containing an 8.6 kb upstream fragment (construct I, see Fig. 1) confer the full normal cyclic pattern (Fig. 5). Double hybridisation with the her1 probe shows that it is essentially indistinguishable from the endogenous expression(Fig. 5G,H) indicating that all important elements for in phase cycling are present. However, the transgenic embryos of all 8.6 kb lines show an additional strong expression in the notochord, which is not seen for the her1 gene itself, suggesting that construct I is missing a notochord-specific repressor. Furthermore, the expression in the most posterior domain is more persistent than for the her1 RNA. It is unclear whether this is due to slight differences in the RNA stability, or to differences in the action of the enhancer included in our construct.

Lines containing 3.3 kb upstream sequence (construct II, see Fig. 1) show only a subset of the expression pattern. There is no expression in the posterior PSM and notochord expression is now absent. However, the stripes in the intermediate and anterior PSM are still generated and show cyclic expression(Fig. 6B,D,F). Taking the tip of the tailbud as reference (bar in Fig. 6B,D,F) one can see three different phases. Three stripes are visible in the first phase, with the most posterior stripe close to the tail bud (Fig. 6B). Two stripes are visible in the other phases, with the most posterior stripe either at an intermediate distance to the tailbud (Fig. 6D) or with a large distance to the tailbud(Fig. 6F). Thus, the most posterior stripe behaves in the same way as the anterior border of the dynamic U-shaped domain (compare Fig. 6A,C,E with B,D,F respectively). Formation of the stripes is obviously independent of wave generation in the posterior but must be linked, since budding of the most posterior stripe appears always at the very end of the peak of the oscillating U-shaped domain in the tailbud.

A further shortening of the upstream region to 2.8 kb (construct III in Fig. 1) shows a loss of dynamic stripe expression in the intermediate of the PSM and only a weakly expressed broad domain is visible (Fig. 6G). Lines containing shorter reporter gene constructs (IV-VI)with 2.3, 0.8 and 0.3 kb upstream sequences did not show any expression narrowing down the elements, which specifically drive her1expression, between 2.3 and 8.6 kb upstream of the transcription start.

To analyse whether the reporter gene constructs depend on her1 and her7 regulation in the same way as the endogenous genes, we have used MO knockdown of her1 and her7 in the background of the respective transgenic lines. For lines containing the 8.6 kb construct we find that the stripes in the anterior PSM are indeed disrupted by her1^mo and her7^mo injection, while the expression in the posterior PSM persists(Fig. 7B,C). A similar picture is seen in lines with the 3.3 kb construct. The stripe formation is clearly disrupted and a broad domain persists instead(Fig. 7E,F). These results confirm that the essential elements of the stripe regulation must be included in our constructs. In particular, the binding site(s) for Her7, which regulate the stripes of her1 expression, must be included in construct II(compare Fig. 7D with F).

However, we do not see evidence for residual cycling in construct I-containing lines and the anterior border of the remaining domain also appears to be anteriorly shifted in both lines, when compared to the endogenous pattern in the respective morphants. This may be explained by a slightly higher MO-mediated stability of the GFP mRNA.

Although the analysis of carefully staged embryos has provided a clear indication that her1 expression is dynamic in the PSM(Holley et al., 2000), this can be demonstrated more directly in the transgenic lines by comparing the GFP mRNA expression with the GFP protein expression. Because the GFP protein is much more stable than the GFP mRNA, it persists in all cells in which the mRNA was at least transiently expressed. Accordingly, if the her1expression moves across the PSM in the same way as it was demonstrated for hairy-like genes in chicken(Palmeirim et al., 1997), then we would expect all somites to express GFP protein. The somites and the PSM show GFP fluorescence in the 8.6 kb line, with a fading of the signal towards the oldest somites (Fig. 8A,B). Since the GFP protein derived from the posterior U-shaped domain could cause this effect alone in the 8.6 kb line, we monitored GFP fluorescence in the 3.3 kb line, which lacks the posterior PSM expression. This line also shows continuous fluorescence, apart of the signal in the posterior part of the PSM(Fig. 8C,D). We can therefore conclude that the stripes in the intermediate and anterior PSM also move across all cells, confirming that the stripes seen in the RNA pattern are indeed solely due to a very short half-life of the mRNA.

Although cyclic expression of hairy/E(spl)-related genes is now well known to be an essential component of somite formation in fish, chicken and mouse, the generation of this dynamic expression is still not fully understood. It is, however, clear that the Delta-Notch signalling pathway is required. Effector genes of the Delta-Notch pathway, such as Suppressor of Hairless-related genes(Oka et al., 1995; Sieger et al., 2003), hairy/E(spl)-related genes (discussed above), or lunatic fringe-related genes (Dale et al.,2003) are involved in this process. However, there are interesting species-specific differences in the recruitment of genes for this process(Prince et al., 2001; Leve et al., 2001). This is also reflected in the phylogenetic relationships between the different oscillating hairy/E(spl)-related genes. Zebrafish genes, such as her9, that are most similar to the respective chicken hairy-like genes are not expressed in the zebrafish PSM(Leve et al., 2001) while a possible orthologue of the zebrafish her1 gene is not evident in mouse or chicken. Moreover, although her1 and her7 in zebrafish are closely linked, they are not the result of recent gene duplication, as each is more similar to other genes in zebrafish than they are to each other. An in depth analysis of the relationships of the hairy/E(spl)-related genes in the completely sequenced genomes of Takifugu and Tetraodon yielded 5 chicken hairy-like genes in pufferfish species and it was suggested that there are probably even more such genes in zebrafish (Gajewski and Voolstra, 2002). However, our own preliminary analysis of the zebrafish genome (assembly 06) does not support this, since we find only one copy of the chicken hairy homologues her6 and her9. Interestingly, 3 copies of her4 are present in the zebrafish genome and one of them is known to be expressed in the PSM (Takke et al., 1999). Since only two copies of her4 exist in Takifugu(Gajewski and Voolstra, 2002)it seems that number and copy of the different hairy/E(spl)-related genes is highly variable even in teleost fish.

The MO knockdown results suggest that her1 and her7 act in a common pathway, as they both affect the other's expression, as well as deltaC and deltaD expression. The effect on the latter two genes is almost indistinguishable between her1 and her7. Knockdown of these bHLH proteins disrupts deltaC and deltaDexpression (Fig. 4). Since overexpression of her1 leads to a decreased transcript level of these delta genes (Takke and Campos-Ortega, 1999), the results are consistent with the proposed role of a Her-linked Delta-Notch feedback loop(Holley et al., 2002; Oates and Ho, 2002).

However, different mutual effects are observed for her1 and her7, on each other, as well as on their own expression. Lack of Her1 protein results in a specific loss of the stripes in the intermediate and anterior PSM, while the dynamic expression in the tailbud and posterior PSM of both bHLH genes appears nearly unaffected. In contrast to previous suggestions(Holley et al., 2002; Oates and Ho, 2002), her1 thus acts formally as an activator rather than a repressor, not only on its own but also on her7 transcription. Since the Her1 protein possesses all features of bHLH repressors (see alignment in supplemental data)and experimental evidence also supports this, another yet unidentified component must be postulated to act in intermediate and anterior PSM. Two working models seem most likely: Her1 might repress transcription of another repressor resulting in own-stripe activation, or, a modulator protein might switch the function of Her1.

The role of her7 is more complex. Loss of Her7 protein leads to disruption of her1 and her7 cycling, indicating that Her7 is needed for the wave generation of both genes in the posterior PSM. Only residual tailbud expression of her7 is visible in her7morphants and in intermediate and anterior PSM her7 is very weakly expressed suggesting that stabilisation of this transcript plays a less important role than in the case of her1 in her1 morphants. her1 also loses its dynamic expression upon Her7 knockdown, but is expressed throughout the PSM. This would lead to the conclusion that Her7 plays different roles in regulating the two bHLH genes in intermediate and anterior PSM, while Her1 regulates both genes in a similar manner. Formally,Her7 acts as an interstripe repressor on her1, but as an activator on itself. Again, Her7 displays all features known for bHLH repressors (see alignment in supplemental data)suggesting that Her7-mediated activation is indirect and might involve an unknown component. Whether this component is the same as postulated for Her1-mediated activation remains to be investigated.

The results discussed above suggest that her7 is specifically required to initiate the dynamic expression wave, while her1 is required to carry it further on. This points towards a functional separation of the regulation in the posterior PSM from that in the intermediate and anterior PSM. The analysis of the promoter elements of her1 confirms this.

Our results suggest that there are at least two distinct elements controlling the PSM expression of her1. One mediates a specific activation in the most posterior region of the PSM and the second mediates the expression in the intermediate and anterior PSM. Genetic analysis of various mutants (van Eeden et al.,1998; Holley et al.,2000), as well as additional experiments(Holley et al., 2002) suggest a threephase model for the activation and action of her1. The first phase is activation through deltaD and deltaC in the most posterior part of the PSM. The second is the generation of the dynamical stripe pattern and the third is the stabilisation of the stripes during the early stages of somite boundary formation(Holley et al., 2002). Our transgenic lines provide support for at least the first two of these phases. A possible enhancer that is required for the activation of the cycles could be located in the region between -8.6 to -3.3 kb. This would explain the absence of the most posterior expression of GFP mRNA in transgenic embryos containing the 3.3 kb construct. We note, however, that this enhancer will still have to be better defined, because the respective construct leads in addition to an ectopic expression in the notochord. The fact that the 3.3 kb construct specifically drives the dynamic expression of the stripes supports the notion that the second phase of expression is driven by a separate enhancer, which includes activating and repressing subelements. The presence of specific activator elements is suggested by the GFP mRNA expression pattern in the line containing the 2.8 kb construct, which shows only a broad domain, but no distinct stripes in the respective region (see Fig. 1). A distinct enhancer for the most anterior stripe, and thus evidence for the third phase of expression, was not detected in our experiments.

Our promoter studies confirm the notion that the dynamic expression of the her1 gene is caused by differential transcriptional regulation,rather than differential mRNA stability, a result that is in line with comparable experiments on lunatic fringe regulation in the mouse(Cole et al., 2002; Morales et al., 2002). However, it is clear that the her1 mRNA must be very unstable, since it would otherwise accumulate in the PSM, like the stable GFP reporter protein. It is thus likely that there is a specific element in the mRNA that causes this instability. A specific 3′-UTR degradation signal has been identified for the gene xhairy2a in Xenopus(Davis et al., 2001). A sequence of 25 bases in the 3′-UTR of this gene seems to be necessary and sufficient for the rapid turnover. Similarities with this 25 base block are found in some other hairy-related mRNAs, like c-hairy1and human Hes4. However, we do not find this motif in the 3′-or 5′-UTRs of her1 and her7. Interestingly, two copies of the motif occur in the 3′-UTR of her9, a gene that is closely related to the c-hairy1 gene, but that is not expressed in the PSM of zebrafish (Leve et al.,2001).

The mRNA of our GFP reporter gene appears to be equally unstable as the endogenous her1 mRNA. This reporter gene contains only the 5′-UTR of her1, suggesting that the destabilisation signal must be located there. Also the fact that the her1 mRNA is stabilised by the her1^mo that binds in the 5′ region of the mRNA would suggest that a destabilisation signal is linked to the 5′-UTR. MOs against her7 5′ mRNA regions stabilise coinjected her7mRNA as well as a GFP hybrid mRNA that contained only the her75′-UTR (Oates and Ho,2002) indicating that the destabilisation signal for her7could also be linked to the 5′-UTR. In contrast, the stabilisation effect of MOs on the endogenous her7 mRNA is apparently only weak, at least when compared to her1 (see Fig. 4).

Holley et al. (Holley et al.,2002) and Oates and Ho (Oates and Ho, 2002) have proposed models that integrate the effects of Delta-Notch signalling in the regulation of her1 and her7 in the PSM. A core component of these models is that the Her proteins repress their own transcription as well as the transcription of the deltagenes, leading to oscillatory gene expression. Our data suggest that her1 and her7 do not act as repressors, but formally as activators. However, since both proteins belong to a family of bHLH proteins,which are only known as repressors, we have to postulate additional components(see above).

Lewis (Lewis, 2003) has proposed a model in which autoinhibition of her1 and her7coupled with transcriptional delay could serve as the basis of an intracellular oscillation, which would be brought into phase by Delta-Notch signalling (Jiang et al.,2000). This model is also not fully in line with our findings,because it requires a direct repression effect of both genes on themselves and a fully equivalent role of both genes. However, the model seems sufficiently flexible to allow for indirect repression and for the possibility of slightly different roles of her1 and her7.

It is clear that any model that takes only Delta-Notch signalling components into account cannot be complete. Holley et al.(Holley et al., 2002) have shown that fused somites (fss) and beamter(bea) are involved in gene regulation in the anterior PSM. Furthermore it is known that an FGF gradient is required for coordinating the segmentation process and the differentiation of cells in the anterior PSM(Dubrulle et al., 2001; Sawada et al., 2001). Finally,results from the mouse indicate that the wnt-signalling pathway could be an essential component that is linked to the Delta-Notch pathway genes(Aulehla et al., 2003).

We want to thank Jose Campos-Ortega for providing us the deltaDand her1 plasmids and Julian Lewis for the deltaC plasmid,as well as communicating the details of his model prior to publication. We also wish to thank Irene Steinfartz for excellent technical assistance. The work was supported by the Deutsche Forschungsgemeinschaft (SFB 572) and by the Fond der chemischen Industrie.